Gauss's Hidden Menagerie: From
Total Page:16
File Type:pdf, Size:1020Kb
Gauss’s Hidden Menagerie: From Cyclotomy to Supercharacters Stephan Ramon Garcia, Trevor Hyde, and Bob Lutz t the age of eighteen, Gauss established the constructibility of the 17-gon, a result that had eluded mathematicians for two millennia. At the heart of his Aargument was a keen study of cer- tain sums of complex exponentials, known now as Gaussian periods. These sums play starring roles in applications both classical and modern, including Kummer’s development of arithmetic in the cyclotomic integers [28] and the optimized AKS primality test of H. W. Lenstra and C. Pomer- ance [1, 32]. In a poetic twist, this recent application of Gaussian periods realizes “Gauss’s dream” of an efficient algorithm for distinguishing prime (a) n = 29 · 109 · 113, ! = 8862, c = 113 numbers from composites [24]. We seek here to study Gaussian periods from a graphical perspective. It turns out that these clas- sical objects, when viewed appropriately, exhibit a dazzling and eclectic host of visual qualities. Some images contain discretized versions of familiar shapes, while others resemble natural phenomena. Many can be colorized to isolate certain features; for details, see “Cyclic Supercharacters.” Historical Context The problem of constructing a regular polygon with compass and straight-edge dates back to (b) n = 37 · 97 · 113, ! = 5507, c = 113 ancient times. Descartes and others knew that with only these tools on hand, the motivated geometer Figure 1. Eye and jewel—images of cyclic supercharacters correspond to sets of Gaussian could draw, in principle, any segment whose periods. For notation and terminology, see length could be written as a finite composition “Cyclic Supercharacters.” Stephan Ramon Garcia is associate professor of mathe- matics at Pomona College. His email address is Stephan. [email protected]. of sums, products, and square roots of rational Trevor Hyde is a graduate student at the University of Michi- gan. His email address is [email protected]. numbers [18]. Gauss’s construction of the 17-gon Bob Lutz is a graduate student at the University of Michi- relied on showing that gan. His email address is [email protected]. All article 2π p q p figures are courtesy of Bob Lutz. 16 cos = −1 + 17 + 34 − 2 17 Partially supported by National Science Foundation Grant 17 r DMS-1265973. p q p q p + + − − − + DOI: http://dx.doi.org/10.1090/noti1269 2 17 3 17 34 2 17 2 34 2 17 878 Notices of the AMS Volume 62, Number 8 (a) n = 3 · 5 · 17 · 29 · 37], ! = 184747, c = 3 · 17 (b) n = 13 · 127 · 199, ! = 6077, c = 13 Figure 2. Disco ball and loudspeaker—images of cyclic supercharacters correspond to sets of Gaussian periods. For notation and terminology, see “Cyclic Supercharacters.” was such a length. After reducing the constructibil- wrote that “[c]yclotomy is to be regarded … as 2π the natural and inherent centre and core of the ity of the n-gon to drawing the length cos n , his result followed easily. So proud was Gauss of this arithmetic of the future” [39]. Two of Kummer’s discovery that he wrote about it throughout his most significant achievements depended critically career, purportedly requesting a 17-gon in place on his study of Gaussian periods: Gauss’s work of his epitaph.1 While this appeal went unfulfilled, laid the foundation for the proof of Fermat’s Last sculptor Fritz Schaper did include a 17-pointed star Theorem in the case of regular primes and later at the base of a monument to Gauss in Brunswick, for Kummer’s celebrated reciprocity law. where the latter was born [31]. This success inspired Kummer to generalize Gauss went on to demonstrate that a regular Gaussian periods in [30] to the case of composite n-gon is constructible if Euler’s totient '(n) is moduli. Essential to his work was a study of d a power of 2. He stopped short of proving that the polynomial x − 1 by his former student, these are the only cases of constructibility; this L. Kronecker, whom Kummer continued to mentor remained unsettled until J. Petersen completed a for the better part of both men’s careers [27]. Just as largely neglected argument of P. Wantzel nearly Gaussian periods for prime moduli had given rise three quarters of a century later [33]. Nonetheless, to various families of difference sets [7], Kummer’s the chapter containing Gauss’s proof has persisted composite cyclotomy has been used to explain deservedly as perhaps the most well-known section certain difference sets arising in finite projective of his Disquisitiones Arithmeticae. Without the geometry [14]. Shortly after Kummer’s publication, language of abstract algebra, Gauss initiated the L. Fuchs presented a result in [23] concerning the study of cyclotomy, literally “circle cutting,” from vanishing of Kummer’s periods that has appeared an algebraic point of view. in several applications by K. Mahler [34], [35]. A The main ingredient in Gauss’s argument is modern treatment of Fuchs’s result and a further an exponential sum known as a Gaussian period. generalization of Gaussian periods can be found Denoting the cardinality of a set S by jSj, if p is in [21]. an odd prime number and ! has order d in the For a positive integer n and positive divisor unit group (Z=pZ)×, then the d-nomial Gaussian d of '(n), Kummer “defined” a d-nomial period periods modulo p are the complex numbers modulo n to be the sum d−1 j ! d−1 j ! X ! y X ! y (1) e ; e ; n p j=0 j=0 × where y belongs to Z=pZ and e(θ) denotes where ! has order d in the unit group (Z=nZ) and exp(2πiθ) for real θ. Following its appearance in y ranges over Z=nZ. Unlike the case of prime moduli, Disquisitiones, Gauss’s cyclotomy drew the atten- however, there is no guarantee that a generator ! tion of other mathematicians who saw its potential of order d will exist or that a subgroup of order × use in their own work. In 1879, J. J. Sylvester d will be unique. For example, consider (Z=8Z) , which contains no element of order 4, as well as 1H. Weber makes a footnote of this anecdote in [43, p. 362], three distinct subgroups of order 2. A similar lack but omits it curiously from later editions. of specificity pervaded some of Kummer’s other September 2015 Notices of the AMS 879 definitions, including his introduction of ideal We are now in a position to clarify the captions prime factors, used to prove a weak form of prime and colorizations of the numerous figures. Unless factorization for cyclotomic integers. According specified otherwise, the image appearing in each to H. M. Edwards, instead of revealing deficiencies figure is the image in C of the cyclic supercharacter × in Kummer’s work, these examples suggest “the σh!i : Z=nZ ! C, where ! belongs to (Z=nZ) , and mathematical culture…as Kummer saw it” [19]. h!i = h!i1 denotes the orbit of 1 under the action Fortunately, the ambiguity in Kummer’s defi- of the subgroup generated by !. Conveniently, nition is easily resolved. For n as above and an the image of any cyclic supercharacter is a scaled × element ! of (Z=nZ) , we define the Gaussian subset of the image of one having the form σh!i [17, periods generated by ! modulo n to be the sums Proposition 2.2], so a restriction of our attention in (1), where d is the order of ! and y ranges to orbits of 1 is natural. Moreover, the image of over Z=nZ, as before. These periods are closely σh!i is the set of Gaussian periods generated by related to Gauss sums, another type of exponential ! modulo n, bringing classical relevance to these sum [9]. figures. To colorize each image, we fix a proper divisor Cyclic Supercharacters c of n and assign a color to each of the layers, In 2008, P. Diaconis and I. M. Isaacs introduced σ (y) j y ≡ j (mod c) ; the theory of supercharacters axiomatically [15], h!i1 building upon seminal work of C. André on the rep- for j = 0; 1; : : : ; c − 1. Different choices of c result resentation theory of unipotent matrix groups [3], in different “layerings.” For many images, certain [4]. Supercharacter techniques have been used to values of c yield colorizations that separate distinct study the Hopf algebra of symmetric functions graphical components. of noncommuting variables [2], random walks on upper triangular matrices [5], combinatorial prop- erties of Schur rings [16], [40], [41], and Ramanujan sums [22]. To make an important definition, we divert briefly to the character theory of finite groups. Let G be a finite group with identity 0, K a partition of G, and X a partition of the set of irreducible characters of G. The ordered pair (X; K) is called a supercharacter theory for G if f0g 2 K, jXj = jKj, and for each X 2 X, the function X σX = χ(0)χ χ2X is constant on each K 2 K. The functions σX : G ! C are called supercharacters, and the elements of K are called superclasses. (a) n = 13 · 127 · 199, X = ! = 9247, c = 127 Since Z=nZ is abelian, its irreducible characters are group homomorphisms Z=nZ ! C×. Namely, for each x in Z=nZ, there is an irreducible character xy χx of Z=nZ given by χx(y) = e( n ). For a subgroup × Ð of (Z=nZ) , let K denote the partition of Z=nZ arising from the action a · x = ax of Ð . The action a · χx = χa−1x of Ð on the irreducible characters of Z=nZ yields a compatible partition X making (X; K) a supercharacter theory on Z=nZ.