Neutral Geometry and the Independence of the Parallel Postulate

S. Weston Muniak May 10, 2012

Abstract For most high school geometry students, the parallel postulate is taken to be true without question. However, for years, geometers have questioned whether or not the aforementioned postulate, Euclids fth postulate, is truly a postulate, and they have attempted to prove it as a theorem. This paper looks at results found in neutral geometry, a geometry that is independent of the parallel postulate, discusses the history behind mathematicians attempts to prove the parallel postulate, and provides rationale through a metamathematical theorem and its corollary as to why the fth postulate is independent of the others and, therefore, cannot be proven using the other four postulates.

Acknowledgements Thanks to Dr. Dale McIntyre, my advisor in this project as well as a dear friend. His enthusiasm for the subject served as a model for the geometry of this paper. Thanks to Dr. Kelly Bonomo, who sparked my initial love for geometry in Math for Secondary Education, giving me and my class the training we needed to form an appreciation for geometry. Thanks to Mr. Dan Dean, who has been a friend and mentor, as well as a respected and beloved professor, for three years of my college career. His work in his students lives has not gone unnoticed by me or my graduating class, and we all have grown to appreciate his impact on our lives as people and as mathematicians. This paper is in memory of Stanley E. Scionka, who passed away March 19, 2012. I always say that he taught me everything I knowby this I mean that he sparked my learning from a very young age, as he was my rst and greatest teacher.

Introduction
Perhaps the greatest mind in the history of geometry, Euclid of Alexandria (. 300 BC) published his magnum opus, the Elements, with the distinction of being the rst to standardize a synthetic geometry, a geometry based purely on axioms and constructions. The Elements remains one of the most inuential works in geometry, as its timeless foundations have inspired countless authors in their quest for geometric truth. Euclid divided his work into sectionsthe rst, denitions and undenable terms; the second, common notions, or axioms; the third, the postulates of geometry; and the fourth, his propositions regarding the axioms and postulates. For our study, the postulates are of specic importance, as they are, by denition, those statements specic to geometry without need of proof. The following is a list of Euclids ve postulates: 1. A straight line segment can be drawn joining any two points. 2. Any straight line segment can be extended indenitely in a straight line. 3. Given any straight line segment, a circle can be drawn having the segment as radius and one endpoint as center. 4. All right angles are congruent. 5. If two lines are drawn which intersect a third in such a way that the sum of the inner angles on one side is less than two right angles, then the two lines inevitably must intersect each other on that side if extended far enough. As Euclid was interested in a geometry developed synthetically, not analytically, points specically located in space did not matter, so each of the rst three postulates is necessarily independent from the others to develop curves and lines in the geometry. The fourth postulate regarding right angles was proved as a theorem in Hilberts expansion and modernization of Euclids work. The fth postulate, at rst glance, is not clearly a postulate at allin fact, many mathematicians have attempted its proof based on the other four postulates. This postulate was proved over time to be equivalent to what is known by many names as the parallel postulatethat through any point not on a given line, there exists exactly one line parallel to the given line. The drama of the ambiguity of the nature of the fth postulate was not lost on Euclidin fact, the rst twenty-eight of his propositions used only the rst four postulates; however, Euclid was forced to use the fth postulate in the proof of his twenty-ninth proposition and in many following.

As a result, geometers have long viewed the fth postulate with uncertainty in terms of its independence from the other postulates. This led to two major questions being asked: 1. What can be proved without Euclids fth postulate? 2. Is Euclids fth postulate independent of the other postulates? In the following sections, we will suggest answers to the preceding questions.

Neutral Geometry
Many results have been able to be proved in a geometry that does not use the parallel postulate. We dene the following: Denition 1. A geometry that uses only the rst four postulates of Euclid (and not the fth) is called neutral, or absolute, geometry. Note that this geometry lacks any mention of parallel lines; therefore, any results involving distinction between one or more parallel lines (as the existence of parallel lines can be proved) must not be considered, and as such, the geometry is not complete. However, it is important that the geometry itself is considered, as its results hold in both Euclidean (in which Euclids fth postulate holds) and the later-discovered hyperbolic (in which an alternate parallel postulate holds) geometries. It is important to look at some interesting results in neutral geometry, as they can often be surprising to the mathematician comfortable with the truth of Euclidean geometry. One such surprising result is the Saccheri-Legendre theorem regarding angle sums of triangles. We will rst introduce relevant notation as well as some conjectures that will be taken to be true for the purposes of our study: Notation. For a triangle ABC, the angle sum s( ABC) = mA + mB + mC.

Now we will state the Saccheri-Legendre theorem with its interesting result. Theorem 2 (Saccheri-Legendre). For any triangle ABC, s( ABC) 180 .

In Euclidean geometry, it is among the most commonly-known ideas that the angle sum of a triangle is always 180 . However, this result cannot be known in neutral geometry! We will introduce a relevant axiom and two lemmas to guide us through the proof of this theorem. Diagrams follow the proof of each lemma for better visualization of the lemmas contents. 3

Archimedes Axiom. Let > 0 and let A be an angle with measure . The degree measure of a certain angle can be halved (or doubled) enough times such that the degree measure of the resulting angle is greater than or equal to mA (if doubled) or less than or equal to mA (if halved). Lemma 3. The sum of the measures of any two angles in a triangle is less than 180 . Proof of Lemma 3. Let ABC be a triangle. Without loss of generality, construct point D

such that D AC (Theorem 0.1, see appendix). Then mBAC +mABC < mBAC + mBAD (Theorem 0.2). Since BAC and BAD form a linear pair, by Theorem 0.3, we have mBAC + mBAD = 180 . Thus mBAC + mABC < 180 .

Lemma 4. For any triangle

ABC, there is another triangle

A B C such that:

1. s( ABC) = s( A B C ), and 2. mA
mA 2 .

Proof of Lemma 4. Let ABC be a triangle. Construct D, the midpoint of BC, and con struct E on AD such that A D E and AD DE (Theorem 0.1). = Consider the newly-constructed ACE. For convenience, let BAD = 1, DAC = 2, DCE = 3, and ACD = 4. Note that BD CD, by the denition of the mid= point, BDA CDE (Theorem 0.4), and AD DE (by construction based on Theo= = rem 0.1). Therefore, BDA = CDE, by Theorem 0.5, the Side-Angle-Side congruence

theorem. By Theorem 0.6 (corresponding parts of congruent triangles are congruent), 1 E and B 3. Note that = = s( ABC) = mA + mB + m4 = m1 + m2 + mB + m4. Also, s( ACE) = m2 + mACE + mE = m2 + m3 + m4 + mE. Since 1 E, we can see that s( ABC) = s( ACE). =

Note that mBAC = m1 + m2 = mE + m2. Since angle measure is dened as a positive real number, it is impossible for both E and 2 to have measure greater than
mBAC . 2

Let A be the smaller of the two (or, if they are the same size, an arbitrary one mBAC . 2

of the two). Thus mA Thus we have proved the lemma.

Now we may proceed to prove the theorem. Proof of the Saccheri-Legendre Theorem. Let

ABC be a triangle whose angle sum is more

than 180 . We can say s( ABC) = (180 + ) . According to Lemma 4, there is a triangle A1 B1 C1 with the same angle sum but mA1 1 mA; 2 A2 B2 C2 with the same angle sum but mA2 1 mA1 1 mA; 2 4 A3 B3 C3 with the same angle sum but mA3 1 mA2 1 mA. 2 8 Clearly, if we iterate n times, we derive a triangle An Bn Cn with angle sum (180 +
1 2n mA.

) but which has an angle An whose measure is less than or equal to Archimedes Axiom, there is a large enough value of n such that
1 2n mA

By

< . In that case,

the remaining two angle measures must add up to more than 180 , which is impossible by Lemma 3. Thus there cannot be a triangle with an angle sum greater than 180 , so s( ABC) 180 . Another interesting result, included because of its equivalent nature in Euclidean and analytic geometry and because of its power to qualify neutral geometry as a metric space, is as follows, with the proof left to the reader: Theorem 5 (Triangle Inequality). In any triangle ABC, the length AC is bounded above

and below by the lengths AB and AC; that is, |AB BC| < AC < AB + BC. This and many other curious results occur in neutral geometry; for further enrichment and enjoyment, the reader is encouraged to consider Greenbergs Euclidean and Non-Euclidean Geometries (1974).

The Independence of the Parallel Postulate

Now that we have discovered a few relevant results in neutral geometry, we must consider the other main question at handis the parallel postulate independent of the rest of Euclids postulates? We will begin by listing several statements that have been proven to be logically equivalent to Euclids fth postulate: Claim 6 (Playfairs Axiom). Given a line and a point not on the line, exactly one line parallel to the given line can be drawn through the point. This axiom is the more common representation of the parallel postulate, and in most cases, we will use it to describe the parallel postulate, eschewing Euclids wording in favor of this simpler and more direct idea. Claim 7. Given two parallel lines and m and a line n through , line n also intersects m. This claims wording is useful in that it establishes the basis for contradiction very succinctly when proving the uniqueness of a line parallel to a given line through a point not on the line. Claim 8 (Wallis Postulate). Given any triangle exists a triangle ABC and given any segment DE, there ABC.

DEF (having DE as one of its sides) that is similar to

This claim was born from an attempt to prove Euclids fth postulate, but it can also be shown to be logically equivalent to the parallel postulate. Clearly, any attempted proof of Euclids fth postulate which uses the aforementioned claims as a basis for any argument within the proof is circular and therefore invalid.

Attempted Proofs of the Parallel Postulate

The rst known attempt at proving the fth postulate as a theorem was in the second century, by Ptolemy. Very little has been reported about his proof, except that in his proof, he assumed Playfairs Axiom, which is logically equivalent to Euclids fth postulate. As a result, Ptolemys argument is invalid, and the fth postulate withstood his initial attack. Not swayed by Ptolemys lack of success in proving the fth postulate, in the fourth century, Proclus gave his own proof of the parallel postulate, interesting in that his foundation was considerable in its strength, and he does not directly violate any of the logicallyequivalent statements to Euclids fth postulate. His proof is as follows (a diagram follows for convenience in understanding): Proclus Proof (False). Given two parallel lines and m, suppose line n cuts m at point P . We want to show that n intersects , as well. 6

 Let Q be the foot of the perpendicular from P to (Theorem 0.7). If n coincides with P Q, then it intersects at Q. Otherwise, one ray P Y of n lies between P Q and a ray P X of m. Take X to be the foot of the perpendicular from Y to m. Now, as the point Y recedes endlessly from P on n, the segment XY increases indenitely in size, and in so doing eventually becomes greater than segment P Q. Therefore, Y must cross over to the other side of , so that n intersects .

Proclus error was that he failed to justify two important assumptions he made. First, he assumed that X, Y, and Z were collinear; he also assumed that XQ P Q. While both = are very reasonable assumptions in Euclidean geometry, there is no denition in neutral geometry that states that parallel lines are uniform in their distance from one another in fact, in hyperbolic geometry, this is denitely not the case! As a result of these faulty assumptions, Proclus proof, while much better than Ptolemys before him, is invalid and cannot be taken as true. It is intuitive that parallel lines are a constant distance aparthowever, intuition based on the Euclidean plane (or any model for any geometry) is not the basis for any geometry the axiomatic method is, and any model must adhere to that method, as opposed to the other way around. Therefore, it is important to rid our thinking of the Euclidean plane when we discuss proving the parallel postulate; since the parallel postulate is crucial to the development of the model that is the Euclidean plane, we cannot assume that the Euclidean plane exists before we have proven the parallel postulate. Thus we must accept that parallel lines are, by denition, lines which never intersecttherefore, we cannot take more than that for granted, including the possibility that parallel lines may be a constant distance from one another. Giovanni Girolamo Saccheri (1667-1733), who used ideas based off Omar Khayyams Discussion on the Difculties in Euclid, also attempted to prove the parallel postulate by listing the potential alternatives (including the parallel postulate) and performing a reductio ad absurdum argument. Saccheris trichotomy was as follows. Through any line and a point not on the line, through that point:

1. exists exactly one parallel line; 2. exist no parallel lines; or 3. exist more than one parallel line. Saccheri was able to nd fault in (2) using Euclids second postulate; however, he had much trouble nding an argument against (3), so as to reduce the possibility to (1) alone. In his arguments, he developed the idea of the Saccheri quadrilateral, a quadrilateral whose base angles are right angles and whose sides adjacent to yet not included between the right angles are congruent (see the end of this section for a diagram of a Saccheri quadrilateral). With (1), the summit angles are also right angles, and with (3), the summit angles were found to be acute. This idea disgusted Saccheri, as he exclaimed that (3) must not be true because it was repugnant to the nature of a straight line! While the high tension and emotion in Saccheris argument was hard to ignore, his proof was far from rigorous, and in fact, Saccheri had just discussed a perfectly valid result in hyperbolic geometry. Therefore, Saccheris attempt to prove the parallel postulate was, as others before him, thwarted by the yet-to-be-discovered existence of hyperbolic geometry, which is consistent with the rst four postulates of Euclid. As a result, Saccheris work Euclid Freed of Every Flaw is considered an early work in non-Euclidean geometry, despite its self-knowledge of its discoveries. Throughout the centuries, many more proofs were attempted for the parallel postulate, including those by Bolyai, Lambert, and Legendre; however, none came to fruition due to their inevitable run-ins with logically-equivalent statements to Euclids fth postulate. Thus the focus shifted to showing the independence of the parallel postulate as opposed to its dependence on other postulates, a marked deferment to Euclids original idea that the parallel postulate was, in fact, a postulate in his geometryhowever, what Euclid did not expect was that the fth postulate could also be replaced by another postulate in order to create a completely different geomety, the geometry on a hyperbolic plane.

The Independence of the Parallel Postulate

The independence of the parallel postulate could not be proved in a purely mathematical sense, and Greenberg refers to the theorem on which the question relies as a metamathematical theorem, one that is beyond the scope of mathematics. We will state this theorem, assuming its truth. Fact 9. If Euclidean geometry is consistent, so is hyperbolic geometry. The proof of this theorem was dependent on nding a consistent model for hyperbolic geometry that was workable in all cases. Various models were found by Beltrami, Klein, and Poincar , among others, and the essential truth of this statement has been veried e over time. Fact 9 leads to an interesting corollary (that is, the question at handthe verication of the independence of the parallel postulate), as it essentially removes the fth postulate from being a result of the rst four (as both Euclids fth postulate and the hyperbolic parallel postulate can be assumed in addition to the other postulates of Euclid without any contradiction either way). Corollary 10. If Euclidean geometry is consistent, then no proof or disproof of the parallel postulate from the rest of the postulates will ever be found, i.e., the parallel postulate is independent of the other postulates. While Fact 9 is a metamathematical theorem, Corollary 10 has a strong logical basis and can be proved. The proof, an indirect argument, is as follows: Proof of Corollary 10. Assume on the contrary that a proof of the parallel postulate exists. Then hyperbolic geometry would be inconsistent, since the hyperbolic parallel postulate contradicts a proved result. But the metamathematical theorem asserts that hyperbolic geometry is consistent relative to Euclidean geometry. This contradiction proves that no proof of the parallel postulate exists. The hypothesis that Euclidean geometry is consistent ensures that no disproof exists either.

Conclusion
As a result of Corollary 10, we can denitively say that Euclids fth postulate is independent of the other four. As a result, nothing can be proved or disproved regarding parallel lines using only the rst four postulates; therefore, neutral geometry is not equivalent to Euclidean geometry. Additionally, Euclidean geometry and hyperbolic geometry can coexist while sharing the rst four of Euclids postulates and differing on the fth. The 9

usefulness of neutral geometry, then, is the results that come from both Euclidean and hyperbolic geometryresults that are true independent of the planes on which the geometries are valid. Further study in this topic could consider another neutral geometry nding results common to Euclidean geometry and one of many other geometries, such as elliptic, Riemannian, or other Archimedean geometries.

10

Reference
Greenberg, Marvin Jay. Euclidean and Non-Euclidean Geometries: Development and History (1974). San Francisco: W.H. Freeman.

11

Appendix: Prerequisite Theorems

0.1 There exists exactly one point P on AB such that AP = x, where x is a positive real number. 0.2 (Exterior Angle Theorem) The measure of an exterior angle of a triangle is greater than the measures of either of its remote interior angles. 0.3 (Linear Pair Theorem) Two angles which form a linear pair are supplementary; that is, the sum of their measures is 180 . 0.4 (Vertical Angles Theorem) Vertical angles are congruent. 0.5 (Side-Angle-Side Congruence Theorem) If two sets of corresponding sides of triangles and theire corresponding included angles are congruent, then the triangles are congruent. 0.6 (CPCTC) Corresponding parts of congruent triangles are congruent. 0.7 The perpendicular dropped from point P not on a line to is unique (and the point at which the perpendicular intersects is called its foot).

12