Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Noether's theorem wikipedia , lookup
Pythagorean theorem wikipedia , lookup
System of polynomial equations wikipedia , lookup
History of trigonometry wikipedia , lookup
Euclidean geometry wikipedia , lookup
Trigonometric functions wikipedia , lookup
Integer triangle wikipedia , lookup
ON ANGLES WHOSE SQUARED TRIGONOMETRIC FUNCTIONS ARE RATIONAL by John H. Conway1 * Charles Radin2 ** and Lorenzo Sadun3 *** 1 Department of Mathematics, Princeton University, Princeton, NJ 08544 2;3 Department of Mathematics, University of Texas, Austin, TX 78712 Abstract We consider the rational linear relations between real numbers whose squared trigonometric values are rational, angles we call \geodetic". We construct a convenient basis for the vector space over Q generated by these angles. Geodetic angles, and rational linear combinations of geodetic angles, appear naturally in Euclidean geometry; for illustration we apply our results to equidecomposability of polyhedra. * Research supported in part by NSF Grant No. DMS-9701444 ** Research supported in part by NSF Grant No. DMS-9531584 and Texas ARP Grant 003658-152 *** Research supported in part by NSF Grant No. DMS-9626698 and Texas ARP Grant 003658-152 0. Introduction Many well known geometric objects involve angles that are irrational when measured in degrees or are irrational multiples of in radian measure. For instance we might mention the dihedral angle ( 70310 4400) of the regular tetrahedron, whose supplement ( 1092801600 ) is known to chemists as the carbon valence bond angle. A goodly number of these angles have the property that their six trigonometric functions have rational squares. For instance, sin2 = 98 ; cos2 = 91 ; tan2 = 8; cot2 = 81 ; sec2 = 9; csc2 = 89 1) and, for the dihedral angle ( 144440 800 ) of the cuboctahedron, sin2 = 32 ; cos2 = 13 ; tan2 = 2; cot2 = 21 ; sec2 = 3; csc2 = 23 : 2) There are many additive relations between angles of this kind; for instance and satisfy + =2 = . In this paper we nd all relations between these angles. To be precise, we shall say that is a \pure geodetic angle" if any one (and therefore all) of its six squared trigonometric functions is rational (or innite), and use the term \mixed geodetic angle" to mean a linear combination of pure geodetic angles with rational coecients. The mixed geodetic angles form a vector space over the rationals and we shall nd an explicit basis for this space. Another aim of this paper is to introduce an elegant notation for these angles which we hope will nd general acceptance. Namely, if 0 r 1 is rational we dene 6 r = sin?1 pr: 3) (We feel free to write these angles in either degrees or radians.) The well-known particular cases are: 6 0 = 0 = 0; 6 1 = 30 = ; 6 1 = 45 = ; 6 3 = 60 = ; 6 1 = 90 = : 4) 4 6 2 4 4 3 We extend this notation by writing, for all integers n, 6 n = 90n = n ; 6 (n + r) = 6 n + 6 r: 2 2 5) Our basis contains certain angles hpid for prime p and square-free d. (Some exercise in the notation is provided in Tables 1 and 2, which show the rst few elements in the basis.) hpid is dened just when ?d is congruent to a square modulo p and is found as follows. Express 4ps as a2 + db2 for the smallest possible positive s. Then 1 s s r 2 2 1 a 1 ?1 ?1 db : = cos = tan 6) s s s 4p s 4p s 4p s a2 The expression is unique except when d = 1 or 3, when we make it so by demanding that b be even (if d = 3) or divisible by four (if d = 1). Our main result is then Theorem 1. Every pure geodetic angle is uniquely expressible as a rational multiple of plus an integral linear combination of the angles hpid . So the angles hpid , supplemented by (or 6 1 or 1), form a basis for the space of mixed geodetic angles. It is easy to nd the representation of any pure geodetic angle in terms of the basis. p Theorem 2. If tan = ab d for integers a; b; d, with d square-free and with a and b relatively prime, and if the prime factorization of a2 + db2 is p1 p2 pn (including multiplicity), then we have = t hp1id hp2 id hpn id 7) for some rational p t. (In fact the denominator of t will be a divisor of the class number of Q ( ?d).) p For example, for tan = 45 3 we nd a2 + db2 = 16 + 75 = 91 = 7 13 and indeed = n ? h7i3 ? h13i3. Our results have many applications but we shall only mention one outstanding one here. In 1901 Dehn [Deh] solved Hilbert's 3rd problem by giving a necessary condition for the mutual equidecomposability of polyhedra in terms of their dihedral angles. In 1965, Sydler proved [Syd] that Dehn's criterion was also sucient. For polyhedra with geodetic dihedral angles our Theorem 1 makes the Dehn-Sydler criterion eective. At the end of this paper we shall apply our theory to the nonsnub Archimedean polyhedra (whose dihedral angles are all geodetic.) 2 hpid = 1 6 dbs = 1 sin?1 db2 1. Angles with polyquadratic tangents and the Splitting Theorem. The addition formula for tangents enables us to show that the tangent of any sum ofp purepgeodetic p of the p angles is a \polyquadratic number", that is a number form ap+ b + c + , with a; b; c rational. For instance, if tan = 2=2 and tan = 3=3, then p p p p 2 = 2 + p p 3=3 = 54 2 + 35 3; tan( + ) = 1 ? 2 3=6 8) p 3p 4 tan( ? ) = 5 2 ? 5 3: We now suppose the sum of a number of pure geodetic angles is an integral multiple of ; let us say 1 + 2 + + 1 + 2 + = m, where we have 2 p p chosen the notation so thatpthe tangents p ofp 1; 2; p are in Qp( d1; p ; dn ) and those of 1 ; 2; are in Q ( d1; ; dn ) d, where d 2= Q ( d1; ; dn ). Then tan(1 + 2 + p + 1 + 2 +p ) and tan(1 +p2 + p? 1 ? 2 ? ) will be of d and a ? b d where a; b 2 Q ( d1; ; dn ). But by assumption the form a + b p a + b d = 0, so a = b = 0, from which it follows that 1 + 2 + ? 1 ? 2 ? is also an integral multiple of . Adding and subtracting we deduce that the two subsums 1 + 2 + and 1 + 2 + are integral multiples of =2. Repeating this argument we obtain Theorem 3 (The Splitting Theorem). If the value of a rational linear combination of pure geodetic angles is a rational multiple of then so is the value of its restriction to those angles whose tangents are rational multiples of any given square root. Sketch of proof. We exhibit the argument in the instance that X X X X Aii + Bj j + Ck k + D` ` = E; 9) i j k ` p p p where the tangents of the 's, 's, 's and 's are rational multiples of 1; 2; 3; 6 respectively and the A's, B 's, C 's, D's and E are rational numbers. By clearing denominators, we may suppose that the A's, B 's, C 's, DP 's and E are P all integers. Applying our P argument with d = 3 we see rst that 2( i Ai i + j Bj j ) and P 2( k Ck k +P ` D` ` ) are multiples of . P Applying it again with d = 2 P integral P shows that 4 i Ai i, 4 j Bj j , 4 k Ck k and 4 ` D` ` are integral multiples of as well. We remark that the same method can be used to show that any angle whose tangent is polyquadratic is a mixed geodetic angle. If is an angle whose tangent is polyquadratic, we express 2 as the sum of two angles each of whose tangents are simpler polyquadratic numbers; repeating the process we obtain pas a p niteplinear 6+ 3+ 2+1. combination of pure geodetic angles. For example, suppose tan = p p p 0 0 Let be an angle p with p tan = ? 6 + 3 ?p 2 + 1.p tan() and tan(00) are conjugates in Q ( 2; 3) under the operation 2 ! ? 2. Let = + and = ? 0 . By the addition formula for tangents, p ) + tan(0) = 2 + 2p3 = ?2 + 6 p3; tan( ) = 1tan( ? tan() tan(0 ) 5 +p2 3 p13 13 10) 0) p p 2(1 + 3) ? 2 tan( ) ? tan( 2 p tan() = 1 + tan() tan(0 ) = = 3 6 ? 2 2: 3+2 3 Thus 2 = + is the sum of two angles, each of whose tangents is the sum of two square roots. By a similar reduction 2 and 2 are each the sum of two angles whose tangents are a single square root, i.e. pure geodetic angles, so 4 is the sum 432 96 2592 of four pure geodetic angles. Specically, 4 = 6 (1 + 441 457 ) + 6 457 + 6 457 + 6 4113 . 3 2. Strmer theory and its generalization. The Splitting Theorem reducespthe study of the rational linear relationships between angles of the form tan?1 ( ab d) topthose with a xed d. These angles are the arguments of algebraic integers a + b ?d and their theory is essentially p the factorization theory of numbers in Od , the ring of algebraic integers of Q ( ?d). The method was rst used by C. Strmer [St] (in the case d = 1) who found all the linear relations between the arctangents of rational numbers using the unique factorization of Gaussian integers. (See also [Con].) We recall Strmer's analysis of the case d = 1 and then generalize it to arbitrary d, which will prove Theorems 1 and 2. It is known that the Gaussian integers have unique factorization up to multiplication by the 4 units: 1; ?1; i; ?i. It is also known how each rational prime p factorizes in the Gaussian integers. Namely: 1) if p ?1 (mod 4) then p remains prime; 2) if p +1 (mod 4), p = a2 + b2 is the product of the distinct Gaussian primes a + ib and a ? ib (for uniqueness we choose a odd, b even, both positive); and 3) 2 = ?i(1 + i)2 \ramies", that is to say it is (a unit times) the square of a Gaussian prime. Now let = 1 23 be the prime factorization of a Gaussian integer . Then plainly arg() arg(1)+arg(2)+arg(3)+ (mod 2). So the arguments of Gaussian primes (together with ) span the subspace of mixed geodetic angles generated by the pure geodetic angles with rational tangent. However, 1) If p = 4k ? 1; arg(p) = 0 and can be ignored 2) If p = 4k + 1 = a2 + b2, then arg(a + bi) + arg(a ? bi) = 0. We dene hpi1 = arg(a + bi) = ? arg(a ? bi). 3) arg(1 + i) = =4. 4) The arguments of the units are multiples of =2, and so are multiples of arg(1 + i). Thus the argument of any Gaussian integer is an integral linear combination of =4 and the angles hpi1, with p 1 (mod 4). It is also easy to see that the numbers and the hpi1 are rationally independent. Otherwise some integral linear combination of them would be an integral multiple of . But suppose for instance that 2hp1i1 ? 3hp2 i1 + 5hp3i1 = 0. The left hand side is the argument of 12 2335 which must therefore be a real number and hence equal to its conjugate 122335. But this contradicts unique factorization of Gaussian integers. The analogue of the Strmer theory for the general case is complicated by the fact that elements of Od may not have unique factorization. However the ideals do. Instead of assigning arguments to numbers, we simply assign an angle to each ideal 4 I by the rule ) if I = () is principal; 11) arg(I) = arg( 1 arg(Is) if I is not principal, s where s is the smallest exponent for which Is is a principal ideal, and () denotes the principal ideal generated by . Recall that for every d the ideal class group is nite, so such an s exists for every ideal, and s divides the class number of Od . Since the generator of a principal ideal is dened up to multiplication by a unit, the argument of an ideal must be understood to be modulo the argument of a unit divided by the class number of Od . It is not hard to check that, for general ideals I and J, arg(IJ) = arg(I) + arg(J); 12) modulo the ambiguity, by a rational multiple of , in the denition of arg. Thus the argument of any ideal (and in particular the principal ideal generated by any algebraic integer) is an integral linear combination of the arguments of the prime ideals. What remains is to determine the nontrivial arguments of prime ideals. As in the case d = 1, there will be one such angle for each rational prime p for which (p) splits as the product of distinct ideals. We illustrate the procedure by working in O5 , for which the ideal factorizations of the rst few rational primes are: p (2) =(2; 1 + ?5)2 p p (3) =(3; 1 + ?5)(3; 1 ? ?5) p (5) =( ?5)2 p p (7) =(7; 3 + ?5)(7; 3 ? ?5) (11) =(11) 13) (13) =(13) (17) =(17) (19) =(19) p p (23) =(23; 22 + 3 ?5)(23; 22 ? 3 ?5) p p (29) =(3 + 2 ?5)(3 ? 2 ?5): Here (x; y) denotes the ideal generated by x and y. The reader will see that (2) ramies as the square of a non-principal ideal and (5) as the square of a principal ideal, (3), (7) and (23) split into products of non-principal ideals, (29) splits as the product of distinct principal ideals, while (11), (13), (17) and (19) remain prime. As in the Strmer case the principal ideals generated by rational primes that remain prime have argument zero and can be ignored. We also ignore those that ramify, since their angles will be rational multiples of . Otherwise we dene hpid to be the argument of one of the two ideal factors of (p), making it unique by requiring 0 < hpid < =2 if the factors of (p) are principal and 0 < hpid < =4 if not. 5 To illustrate this we determine h3i5. The ideal factors of (3) are non-principal so we square them: p p p (3; 1 + ?5)2 = (32; 3(1 + ?5); (1 + ?5)2 ) p p = (9; 3 + 3 ?5; ?4 + 2 ?5); p p 14) p which reduces to (2 ? ?5). Similarly, (3; 1 ? ?5)2 = (2 + ?5). So h3i5 = 1 arg(2 + p?5) = 1 tan?1 ( 1 p5) = 1 6 5 . 2 2 2 2 9 In the general case (d an arbitrary square-free positive integer) we assign an angle hpid to every rational prime p for which (p) splits as the product of two distinct prime ideals I and J. Let s be the smallest integer for which Is (and p therefore Js) is b a principal. Recall that the elements of Od are of the form 2 + 2 ?d, where a and b are rational integers. If d 6= 3 (mod 4), then a and b must be even; if d = 3 (mod 4) then a and b are either both even or both odd. We can therefore write a p a bp b ?d ; s s ? d ; J = 15) + ? I = 2 2 2 2 where we can distinguish p between I and J by supposing that a and b are positive. 2d 1 1 b b ? 1 6 We take hpid = s tan ( a d) = s a2 +b2 d . The above denes hpid uniquely for all d other than 1 and 3, because then the only units are 1, so that the only generators of Is and Js are the four numbers p a2 2b d. When d = 1 we have the additional unit i which eectively allows us to interchange a and b: we then achieve uniqueness by demanding that the generators of I be a + bi with a and b positive integers with b even. In the case d = 3 the eld has six units and p the corresponding condition is that the generators of I should have the form a + b ?3 where a and b are positive integers. Theorem 4. For xed square-free d, consider the subspace of mixed geodetic angles p generated by arctangents of rational multiples of d. This subspace is spanned by the nonzero angles hpid and , where p ranges over the rational primes for which (p) splits as a product of distinct ideals in Od . Proof. The proof is essentially that of the Strmer decomposition, only substituting the arguments of ideals for the arguments of algebraic integers. Combining Theorem 4 with the Splitting Theorem, we obtain Theorem 1. Proof of Theorem 2. Recall that, if I is any ideal (principal or not) in Od , then II is a principal ideal with a positive integer generator that we call the norm of I, and that norms are multiplicative. From this it follows that the prime ideals are precisely the factors of (p), with p ranging over the rational primes, and that every prime ideal that is not generated by p a rational prime has rational prime norm. b Now suppose that tan() = a d with d square-free and with a and p b relatively prime. Consider the factorization of the principal ideal I = (a + b ?d). If I = 12 n , where each ideal i is prime, then none of the i 's are generated by 6 rational primes, insofar as a and b are relatively prime. Thus each i satises i i = (pi ) for some rational prime pi . We then have (p1 )(p2) (pn ) = (a2 + b2d) = II = 11 n n : 16) So, up to permutation of the indices 1; : : :; n, wepmust have i i = (pi ), and so the argument of i is hpi id . But = arg(a + b ?d) = arg(I) is the sum of the arguments of the i 's, up to a rational multiple of that comes from the ambiguity in the denition of the argument of an ideal. All that remains is to identify the pairs (p; q) for which hpid is nonzero. The following theorems give the criteria. These may be easily checked, by hand for small d and p, and by computer for larger values. The theorems themselves are standard results, and we abbreviate the proofs. Theorem 5. Let p be an odd rational prime. The ideal (p) of Od splits as a product of distinct ideals if and only if we can write 4ps = a2 + b2d 17) for integers a and b (neither a multiple of p) and for an exponent s that divides the class number of Od . If d 6= 3 (mod 4), or if d = 3, then the factor of 4 is unnecessary, and the criterion for splitting reduces to ps = a2 + b2d; 18) for a and b nonzero (mod p). The ideal (p) is prime in Od if and only if ?d is not equal to a square modulo p. If (p) is not prime and does not split, then (p) ramies. Proof. Assume that d is not 1 or 3, so that the only units in Od are 1. If (p) = IJ, let s be the order of I in the ideal class group. Then (ps ) = Isp Js, where Is is a principal ideal. Since the elements of Od are all of the form a2 + 2b ?d, with a and b rational integers, we can write p a b p a2 b2 a b s s s s (p ) = (p) = I J = 2 + 2 ?d 2 ? 2 ?d = 4 + 4 d ; 19) which implies that 4ps = a2 + b2 d. Now, if d 6= 3 (mod 4), then the only way to make a2 + b2 d divisible by 4 is to have both a and b even, in which case we cancel the factors of 2pand write pps in the form a2 + b2 d. Conversely, suppose 4ps = a2 + b2d = [a + b ?d][a ? b ?d]. The left hand side is divisible by p, while each factor in pthe right hand side p is not, so p cannot be prime. Since the only units are 1, (a + b ?d) and (a ? b ?d) are distinct ideals, so the factors I and J of (p) must be distinct, so (p) splits. 2 (mod p), where x is a rational integer, then x2 + d = [x + p?d][x ? If ? d = x p p ?d] is divisible by p. However, neither x ?d is divisible by p, so y(pp) cannot be prime. Conversely, if (p) is not prime, then there are elements x2 + 2 ?d and 7 p b by p, whose product is divisible by p. But then 2 + 2 ?d of2Od , 2neither divisible the norms [x + y d]=4 and [a2 + b2 d]=4 multiply to a multiple of p2, so at least one of these norms is divisible by p. Suppose [x2 + y2d]=4 is divisible by p, so x2 = ?y2 d p(mod p). If y = 0 (mod p), then x = 0 (mod p), contradicting the fact that x2 + y2 ?d is not in (p). Since y 6= 0 (mod p), we can nd an integer z such that yz = 1 (mod p), and then ?d = (xz)2 (mod p). The cases d = 1 and d = 3 are complicated slightly by the existence of units other than 1, but the analysis is essentially the same. Theorem 5 gives criteria for all odd primes. The prime p = 2 is somewhat dierent. Since both 0 and 1 are squares, every ?d is congruent to a square modulo 2. However, there are values of d for which (2) is prime. Theorem 6. If d 6= 3 (mod 4), then (2) ramies in Od . If d = 3 (mod 8), then (2) is prime. If d = 7 (mod 8), then (2) splits and we can write a power of 2 as a2 +pbd2 . Proof. If d 6= 3 (mod 4), then the algebraicpintegers all have thepform a +pb ?d, with a and b rational integers. Either ?d = ( ?d)2 or 1+ d = [1+ ?d][1 ? ?d] is divisible by 2, but none of the factors are divisible by 2, so 2 is not prime. However, (2) does not split, for if 4 2s = a2 + db2, with s minimal, then the left hand side is divisible by 4, while the right hand side is only divisible by 4 if both a and b are even, in which case s is not minimal. Thus (2) ramies. (This argument breaks down if d = 1, but then (2) = ([1 + i]=2)2, so the conclusion still holds). If d = 3 (mod 8), then the algebraic integers that are not divisible by 2 are precisely those with odd norm. Since the product of any two such numbers is another such number, (2) is prime. p p If d = 7 (mod 8), then [d + 1]=4 = [ 21 + 21 ?d][ 12 ? 21 ?d] is divisible by 2, while neither factor on the right hand side is, so (2) is not prime. The fact that (2) splits, as opposed to ramifying, is more dicult, and is not proven here. a 3. Applications to the Dehn-Sydler criterion of Archimedean polyhedra. The Dehn invariant of a polyhedron whose ith edge has length `i and dihedral P angle i is the formal expression i `iV [i] where the \vectors" V [i ] are subject to the relations V [r + s] = rV [] + sV []; V [r] = 0; 20) for all rational numbers r and s. The V [] satisfy the same rational linear relations satised by the angles in the rational vector space they generate, together with the additional relation V [] = 0; however we allow their coecients to be arbitrary real numbers. If every dihedral angle of a polyhedron is geodetic we can write = r + r1 hp1 id1 + + rj hpj idj 21) for rational numbers r; r1; ; rj , so V [] = r1 V [hp1id1 ] + + rj V [hpj idj ]: 22) 8 Its Dehn invariant will then be a rational linear combination of the V [hpid ]. Now each face of an Archimedean polyhedron (other than the snub cube and snub dodecahedron) is orthogonal to a rotation axis of one of the Platonic solids, and the rotation groups of all the Platonic solids are contained in the icosahedral group I . It follows that the dihedral angles of all the polyhedra are found among the angles between thep rotation axes of I . Now let = (1+ 5)=2 and = ?1 = ? 1. The 12 vectors whose coordinates are cyclic permutations of 0; 1; lie along the pentad axes. Similarly the 20 vectors obtained by cyclicly permutating 1; 1; 1 and 0; ; lie along the triad axes, and the 30 cyclic permutations of 2; 0; 0 and 1; ; lie along the dyad ones. The cosines of the angles between the axes have the form v w=jvj jwj where v and w are chosen from these vectors. These cosines are enumerated in Fig. 1. The angles that correspond to them are those shown in Fig. 2, together with their supplements. Table 3 gives the components of the Dehn invariants for the non-snub Archimedean polyhedra of edge lengths 1. For instance the dihedral angles of the truncated tetrahedron are ? 2h3i2 at six edges and 2h3i2 at the remaining 12, so that its Dehn invariant is 6V [ ? 2h3i2] + 12V [2h3i2] = 12V [h3i2]; 23) since V [] = 0. In the values we abbreviate V [hpid ] to hpid . We note that the Dehn invariants of the icosahedron, dodecahedron and icosidodecahedron with unit edge lengths, namely 60h3i5; ?30h5i1, and 30h5i1 ? 60h3i5, respectively, have zero sum, so Sydler's theorem shows that it is possible to dissect them into nitely many pieces that can be reassembled to form a large cube. This might make an intriguing wooden puzzle if an explicit dissection could be found. (We have no idea how to do this.) 4. Angles with algebraic trigonometric functions. It is natural to consider a generalization of our theory that gives a basis for the rational vector space generated by all the angles whose six trigonometric functions are algebraic. What is missing here is the analogue of our Splitting Theorem. If such an analogue were found, the ideal theory would probably go through quite easily. We ask a precise question: Does there exist an algorithm that nds all the rational linear relations between a nite number of such angles? The nicest answer would be one giving an explicit basis, analogous to our hpid . 9 References [Con] J.H. Conway, R.K. Guy, The book of numbers, Copernicus, New York, 1996. [Deh] M. Dehn, Uber den Rauminhalt, Math. Ann., 55 (1901), 465-478. [Syd] J.P. Sydler, Conditions necessaires et susantes pour l'equivalence des polyedres l'espace euclidien a trois dimensions, Comm. Math. Helv., 40 (1965), 43-80. [St] C. Strmer: Sur l'application de la theorie des nombres entiers complexes a la solution en nombres rationnels x1; x2 : : : xn c1 c2 : : :cn ; k de l'equation c1 arc tg x1 + c2 arc tg x2 + cn arc tg xn = k 4 , Arch. Math. Naturvid., 19 No. 3 (1896). 10 dnp 2 3 1 0 2 3 0 5 6 7 10 11 0 13 14 15 1 2 17 1 2 0 56 1 81 4 6 21 22 0 23 1 32 3 6 23 27 0 6 19 20 1 2 6 21 25 0 0 1 4 6 6 16 17 6 0 0 18 19 0 0 0 45 49 0 0 0 0 0 0 0 0 0 0 7 23 0 10 19 1 2 13 49 96 121 6 7 11 6 10 11 6 1 2 0 0 0 6 160 169 0 117 121 6 1 4 0 6 1 2 6 19 44 0 21 121 0 0 0 1 2 0 0 0 1 3 6 0 1 2 240 289 6 153 169 6 6 3 19 6 6 1 2 1 4 1 2 0 19 68 6 189 289 88 169 0 828 2197 0 6 6 1 2 1 4 6 1 2 336 361 6 6 6 6 1 2 352 361 6 0 6 1 2 0 0 0 0 0 27 31 12 37 0 11 92 0 0 240 529 0 187425 279841 0 19 23 0 525 529 0 1 3 6 6 1 2 6 792 841 4508 1 24389 3 6 6 6 6 0 5 41 6 1 2 0 18 43 0 27 43 0 405 1 1849 2 0 0 0 28 37 0 7 43 640 1369 40 41 0 99 148 0 0 0 0 0 0 0 0 0 15 31 0 0 0 0 0 0 0 171 172 6 1 2 6 6 0 1 2 32 41 6 0 0 907137 923521 6 16 41 6 47 6 31 637 961 6 6 1 4 36 37 6 43 0 99 124 6 41 0 0 13 29 6 14 23 6 28 29 6 1 2 6 216 841 6 37 4 29 360 529 22 23 31 20 29 6 0 1 2 6 6 15 19 6 6 45 529 6 1 2 24696 130321 6 6 1 2 325 361 6 29 0 0 13 17 6 4536 28561 6 6 0 3825 1 14641 2 6 6 0 1 2 19 28 6 8 17 23 12 13 1377 1 2401 4 6 4 13 6 19 3 7 0 1 2 17 0 40 49 6 13 2 11 6 1 2 17 81 6 11 6 7 6 504 625 6 1 4 6 11 20 6 0 1 4 1 2 0 11 12 6 6 24 25 6 0 0 6 1 2 0 0 1 3 7 0 0 5 9 6 19 23 0 15 16 6 4 5 6 2 3 6 7 8 6 5 336 961 22 31 207 29791 6 6 6 0 21 37 1 2 0 0 6 6 6 6 525 1681 0 1 3 6 23552 68921 2205 2209 6 0 0 1 2 360 2209 6 11 47 6 1 2 1053 2209 6 0 1 2 2160 2209 6 0 19 188 6 0 1 2 6 0 1408 1849 0 6 1 3 6 22 47 103247 103823 Table 1. Basis elements hpid for some p and d. 0 indicates that (p) is prime in Od , while * indicates that (p) ramies. 11 h5i1 = 6 45 63260600 h13i1 = 6 134 334102400 0 00 h17i1 = 6 16 17 75 57 50 h29i1 = 6 294 21480500 0 00 h37i1 = 6 36 37 80 32 16 16 h41i1 = 6 41 383903500 h3i2 = 6 23 54440800 h11i2 = 6 112 251402200 h17i2 = 6 178 431805000 0 00 h19i2 = 6 18 19 76 44 14 30 4200 62 h41i2 = 6 32 41 18 6 h43i2 = 43 401805500 h7i3 = 6 37 405303600 0 00 h13i3 = 6 12 13 73 53 52 3 23 240 4800 6 h19i3 = 19 0 00 h31i3 = 6 27 31 68 56 54 420 5400 h37i3 = 6 12 34 37 27 6 h43i3 = 43 522403900 h3i5 = 12 6 59 2450 4100 0 00 h7i5 = 12 6 45 49 36 41 57 1 6 45 h23i5 = 2 529 8 2804300 0 00 h29i5 = 6 20 29 56 8 44 h41i5 = 6 415 202602200 405 135701000 h43i5 = 12 6 1849 0 00 h47i5 = 12 6 2205 2209 43 46 50 0 00 h5i6 = 21 6 24 25 39 13 53 6 6 h7i6 = 7 67 4703200 96 31 280 5600 h11i6 = 12 6 121 1 216 h29i6 = 2 6 841 15 1303100 h31i6 = 6 316 2660000 h2i7 = 6 78 691704300 h11i7 = 6 117 525404800 h23i7 = 6 237 332805600 0 00 h29i7 = 6 28 29 79 17 54 28 h37i7 = 6 37 602605700 h43i7 = 6 437 234704400 0 00 h7i10 = 12 6 40 49 32 18 42 270 600 h11i10 = 6 10 72 11 0 00 h13i10 = 21 6 160 169 38 19 44 300 3100 h19i10 = 6 10 46 19 0 00 h23i10 = 21 6 360 529 27 47 29 1 6 640 h37i10 = 2 1369 21 340500 0 00 h41i10 = 6 40 41 81 0 54 360 h47i10 = 21 6 2209 11 5401700 0 00 h3i11 = 6 11 12 73 13 17 520 1100 h5i11 = 6 11 47 20 130 4600 h23i11 = 6 11 20 92 99 6 h31i11 = 124 631901100 99 h37i11 = 6 148 545202100 h47i11 = 6 11 28 550 5700 47 h h h h h h h i 7 13 = 12 6 13 15 300 500 49 11 13 = 21 6 117 39 450 4400 121 580 5800 17 13 = 6 13 60 17 19 13 = 21 6 325 35 470 4500 361 10 5200 29 13 = 6 13 42 29 31 13 = 21 6 637 27 150 700 961 1053 1 47 13 = 2 6 2209 21 490 5500 i i i i i i 0 00 h3i14 = 14 6 56 81 14 3 46 580 2700 h5i14 = 14 6 504 15 625 4536 1 6 h13i14 = 4 28561 55201700 24696 h19i14 = 14 6 130321 6 270500 160 4100 h23i14 = 6 14 51 23 0 00 h2i15 = 12 6 15 16 37 45 40 500 3200 h17i15 = 12 6 240 32 289 410 1800 h19i15 = 6 15 62 19 0 00 h23i15 = 21 6 240 529 21 10 17 40 3300 44 h31i15 = 6 15 31 0 00 h47i15 = 21 6 2160 2209 40 43 2 0 00 h3i17 = 41 6 17 81 6 48 59 h7i17 = 41 6 1377 12 180 2400 2401 1 6 3825 h11i17 = 4 14641 7410500 0 00 h13i17 = 12 6 153 169 36 2 24 h23i17 = 14 6 187425 13 430 5100 279841 1 6 907137 h31i17 = 4 923521 203501100 0 00 h5i19 = 6 19 20 77 4 45 270 4500 h7i19 = 6 19 55 28 19 6 h11i19 = 44 4140 5300 0 00 h17i19 = 6 19 68 31 54 38 210 1000 h23i19 = 6 19 65 23 171 6 h43i19 = 172 853703700 19 h47i19 = 6 188 183201100 0 00 h5i21 = 21 6 21 25 33 12 39 1 6 21 h11i21 = 2 121 121803600 0 00 h17i21 = 12 6 189 289 26 59 3 220 1600 h19i21 = 12 6 336 37 361 525 1 6 h23i21 = 2 529 423002100 0 00 h31i21 = 12 6 336 961 18 7 29 530 000 h37i21 = 6 21 48 37 525 h41i21 = 12 6 1681 16 5901700 88 h13i22 = 12 6 169 23503600 0 00 h19i22 = 12 6 352 361 40 27 28 22 6 h23i22 = 23 77 5705300 0 00 h29i22 = 12 6 792 841 38 0 58 230 4900 h31i22 = 6 22 57 31 0 00 h43i22 = 12 6 1408 1849 30 22 59 22 6 h47i22 = 47 43 1001100 0 00 h2i23 = 31 6 23 32 19 19 27 1 23 6 h3i23 = 3 27 222701500 828 h13i23 = 13 6 2197 12 3702800 4508 8 290 1600 h29i23 = 13 6 24389 207 1 350 3800 h31i23 = 13 6 29791 0 00 h41i23 = 13 6 23552 68921 11 55 27 340 3400 h47i23 = 13 6 103247 28 103823 Table 2. The values of some basis elements hpid . 12 Tetrahedron Truncated tetrahedron 6 ? 12h3i2 6 + 12h3i2 Cube Truncated cube 6 30 ? 24h3i2 Octahedron Truncated octahedron Rhombicuboctahedron 24h3i2 24 9 ? 24h3i2 Cuboctahedron Truncated cuboctahedron 24 ? 24h3i2 54 Icosahedron Truncated icosahedron 15 + 60h3i5 60 + 30h5i1 Dodecahedron Truncated dodecahedron Rhombicosidodecahedron ?30h5i1 75 ? 60h3i5 105 + 60h3i5 ? 30h5i1 Icosidodecahedron Truncated icosidodecahedron 45 ? 60h3i5 + 30h5i1 150 Table 3. The Dehn invariant for the non-snub unit edge Archimedean polyhedra. 13 1; 2 ; 12 ; 2 p 0; 35 ; 13 11 1 1 1 11 11 dyad 1 1 1 1 11 11 q q 11 p3 ; 1; p3 ; 0 p5 ; p5 ; 0 11 11 11 11 11 11 11 11 11 triad pentad 1 1 1 1 1 q 3 q 3 3p5 ; 3p5 p 1; 55 Figure 1. Cosines of angles between axes of xed rotational symmetry (shown at corners), and between axes of dierent rotational symmetry (shown at edges). 14 0; 5 ; 3 ; 25 ; 2 11 1 1 1 1 dyad 1 1 1 11 11 11 11 ; + h3i ; h3i ; ? h3i 1 1 21 h5i1 ; 2 ? 12 h5i1 ; 2 5 2 5 2 4 4 11 11 11 11 11 11 11 11 11 pentad 1 1 1 0; 2 ? 2h3i5; triad 11 ? 2h3i2 3 ? h3i5 ? 1 h5i1 ; 1 + h3i5 ? 1 h5i1 4 2 4 2 h5i1; 0 Figure 2. Angles between axes of xed rotational symmetry (shown at corners), and between axes of dierent rotational symmetry (shown at edges). 15