Introducing an Analysis in Finite Fields

Introducing an Analysis in Finite Fields Hélio M de Oliveira and Ricardo M Campello de Souza Universidade Federal de Pernambuco UFPE CP78-57-97 - Recife - PE, Brazil E-mail: {hmo,ricardo}@ufpebr Abstract - Looking forward to introducing an analysis in Galois Fields, discrete functions are considered (such as transcendental ones) and MacLaurin series are derived by Lagrange's Interpolation A new derivative over finite fields is defined which is based on the Hasse Derivative and is referred to as negacyclic Hasse derivative Finite field Taylor series and -adic expansions over GF(p), p prime, are then considered Applications to exponential and trigonometric functions are presented Theses tools can be useful in areas such as coding theory and digital signal processing INTRODUCTION One of the most powerful theories in mathematics is the Real Analysis [] This paper introduces new tools for Finite Fields which are somewhat analogous to classical Analysis Finite fields have long been used in Electrical Engineering [], yielding challenging and exciting practical applications including deep space communication and compact discs (CD) Such a theory is the basis of areas like algebraic coding, and is related, to various degrees, to topics such as cryptography, spread-spectrum and signal analysis by finite field transforms [,4] Some new and unexpected applications of this framework are currently under developing which include bandwidth efficient code division multiplex [5] Hereafter we adopt the symbol := to denote equal by definition, the symbol to denote a congruence mod p and K[x] a polynomial ring over a field K FUNCTIONS OVER GALOIS FIELDS For the sake of simplicity, we consider initially Galois Fields GF(p), where p is an odd prime Functions: Let be an element of GF(p) of order N i) One of the simplest mapping is the "affine function" which corresponds to a straight line on a finite field, eg a(x)=x+4, x GF(5) x= 4 a(x)= 4 There exists p(p-) linear GF(p)-functions ii) Exponential function: i, i=,,,n- eg letting =, the exponential function x corresponds to i (mod 5) i= 4 indexes i mod 5 4 It is also interesting to consider "shortened" functions in which indexes are reduced modulo p-k for <k<p i= 4 indexes i mod 4 4 repeat Notice that (x) x and (i) i Composite functions can also be generated as usual For instance, the function i= 4 4 corresponds to x 4 (mod 5) = x (x 4 ) Therefore inverse functions over GF(p) can be defined as usual, ie, a function a(x) has an inverse a - (x) iff a(x) a - (x)x mod p There are p! inversive GF(p)-functions The inverse of a(x)x+4 is a - (x)x+ mod 5 Since the characteristic of the field is different from, we can define the odd and even component of a function, respectively, as: a (x ) a(x ) o(x ): and a(x ) a (x ) e (x ): An even (respectively odd) function has o(x) (respectively e(x)) The x exponential function defined over GF(5) has x= 4 x= 4 o(x)= e(x)= 4 4 iii) k-trigonometric functions A trigonometry in Finite Fields was recently introduced [6] Let G(q) be the set of Gaussian integers over GF(q),ie G(q):={a+jb, a,b GF(q)}, where q=p r, p being an odd prime, and j =- is a quadratic non-residue in GF(q) With the usual complex number operations, we have then that GI:= <G(q),,> is a field Definition Let have multiplicative order N in GF(q) The GIvalued k-trigonometric functions cos k () and sin k () are: cos k ( i ): ( i k ik ), sin k ( i ): j ( i k ik ) i,k=,,n- The properties of the unit circle hold, ie, sin k (i ) cosk (i ) mod p We consider now k-trigonometric functions with = an element of order 6 in GF(7) A pictorial representation of the cos-function on this finite field is presented in the sequel There exists several k-cos: {cos k (i)}, k=,,,5 In order to clarify the symmetries, half of the elements of GF(p) are considered positive

and the other half assumes negative values Therefore, we represent the Galois field elements as the set p The cos k (i) over GF(7) assumes the following values (k,i=,,5): cos (i)= (mod 7) cos (i)= 4 6 4 - - - (mod 7) cos (i)= (mod 7) cos (i)= 6 6 6 - - - (mod 7) cos 4 (i)= cos(i) cos 5 (i)= 4 6 4 cos(i) - - - - Figure Pictorial representation of cos k (i) in GF(7) - - each carrier (oscillatory behavior) We define on GF(p) as ():=(p-)/ Clearly, cos (i+())=-cos i We shall see that any function from GF(p) to GF(p) defined by N p- points can be written as a polynomial of degree N- (ie, N indeterminates), eg, i i (mod 5) (indexes i mod 4) Generally there are N unknown coefficients, f(i) a +a i+a i ++a N- i N- mod p, and such a decomposition corresponds to the (finite) MacLaurin series with the advantage that there are no errors DETERMINING THE SERIES Given a function f(i), the only polynomial "passing through" all pair of points {i, f(i)} can be found by solving a Linear System over GF(p) [7] For instance, the -cos function over GF(5) leads to: 4 4 a a a a 4 (mod 5) Another way to find out the coefficients of the polynomial expansion is by using Lagrange's interpolation formula [8], (i )(i )(i ) L (i) ( )( )( ) (i )(i )(i ) 4 ( )( )( ) (mod 5) Therefore cos( i) i i i (mod 5) Identically, the -sin function [6] can be expanded: i= sin (i) j j, sin (i) j(i i i) (mod 5) Euler's formula over a Galois field proved in [6] can also be verified in terms of series: cos(i)+jsin(i) ( i i i ) (i i i) i + i mod 5 (i mod 4) Those results hold despite mod 5! The unicity of the series decomposition can be established by: Proposition Given a function f defined by its values f(x), xgf(p), there exist only one Maclaurin series for f Proof Letting f(x) a +a x+a x ++a p- x p- ( mod p) then f()a and we have the following linear system The "dough-nut ring" representation is well suitable for handling with the cyclic structure of finite fields The envelope does not exists: It just gives some insight on the concept of different "frequencies" on

p p p ( p ) ( p ) ( p ) ( p ) a f( ) f( ) a f( ) f( ) a f( ) f( ) (mod p) a p f(p ) f( ) or Va = f Since V is a Vandermonde matrix, its determinant is: det[v ] (x j x i ) (mod p) where x i :=i i j Therefore, det[v ] mod p concluding the proof 4 DERIVATIVES ON FINITE FIELDS Over a finite field, we are concerned essentially with derivative of polynomials since other functions can be expanded in a series We denote by a (i) (x) the i- th derivative of the polynomial a(x) The concept of classical derivative however presents a serious drawback, namely the derivatives of order greater or equal to the characteristic of the field vanish Let N a(x) a i x i where (i) a i GF(p), x is a dummy i variable Then the derivatives a (p) (x) and a (i) (x) (i p) no matter the coefficients A more powerful concept of derivative over a finite field was introduced a long time ago (see [9]): Definition The Hasse derivative of a polynomial N function a(x) a i x i is defined by i a [ r] N i ( x): i r a i x ir with i r for i<r Clearly the classical Newton-Leibnitz derivative yields a ( r) N ( x) i(i ) (i r ) a i x i r It follows i that r! a( r) (x) a [ r] ( x) The Hasse derivative has been successfully used on areas where finite fields play a major role such as coding theory [] In the following, we introduce a new concept of derivative in a finite field In order to exploit its cyclic structure, a polynomial ring GF(p)[x] is required Let be the constant function f(x)= xgf(p) Conventional modulo x p so that d dx Hasse derivative considers polynomial On the other hand, the fact that i i<r does not allow to handle r with negative powers of x In contrast with Hasse derivative which always decreases the degree of the polynomial, we introduce a new concept of negacyclic Hasse derivative considering the polynomial ring GF(p)[x] reduced modulo x p- + [] Thus, the degree of a(x) is deg a(x)=p- The derivative of a constant no longer vanishes and the degree of the polynomial function is preserved! Over GF(7), for instance, we deal with polynomials of degree 5 and assume d d x d (x 6 ) x 5 mod 7 (negacyclic) d x whereas d d x d (x 7 ) mod 7 (Hasse) d x 5 -ADIC EXPANSIONS AND FINITE FIELD TAYLOR SERIES The classical Taylor series of a function a(x) is given by a(x) a(x ) a() ( x )! a (N ) (x ) (x x o ) N (N )! (x x ) a( ) ( x ) (x x )! R n where Rn a (N ) () (x x o )N N! (Lagrange's remainder) (x x ) On a finite field, Taylor series developed around an arbitrary point GF(p) can be considered according to a(x ) a( ) a [ ] ( )(x ) a [] ()(x ) a [p ] ()(x ) p It is interesting that working with x p ie, deg a(x)=p-, we have R p a (p ) () (x ) p (Section 4) and the p! series is finite (no rounded error, although is unknown) Now let the -adic expansion of a(x) be a(x ) b b (x ) b (x ) b N (x ) N with b =a() It follows from Definition that a [r ] N i (x ): r b i x i r so that a [r ] ( ) r r b r i r The unicity of the Taylor Finite Field series follows directly from []: Unicity Let K be a field and K[x] be a ring of polynomials over such a field The adic expansion of a polynomial a(x) over a ring K[x] is unique

Any function a(x) over GF(p) of degree N- can be written, for any element GF(p), as N a(x) a [ i] ()( x ) i without error or i approximations For instance, the x GF()-function can be developed as x x x mod (McLaurin series) Therefore, ( x ) [ ] x ; x [] ( )! mod Letting x b b (x ) b (x ) mod then x (b b b ) (b b )x b x with b = mod The -adic expansions mod p= are: x x x -adic x (x ) -adic x (x ) (x ) -adic Indeed, a(x ) a() x a [] N () a [i ] ()(x ) i, so one i can, by an abuse, interpret the Hasse derivative as a "solution" of a "/-indetermination" over GF(p): a [] ()? a(x ) a () x 6 APPLICATIONS: EXPONENTIAL AND TRIGONOMETRIC FUNCTIONS Let us consider classical (real) series: e x x x! x! cos(x ) x! x 4 4! x 6 6! + sin(x ) x x! x 5 5! x 7 7! + As an example, we consider the ring GF(p)[x] with x p- + so negacyclic Hasse derivatives can be used Since deg a(x)=p- we truncate the series yielding (e x ) 7 x 4 x 6 x 5 x 4 x 5 (mod 7), (cos x ) 7 4 x 5 x 4 (mod 7), (sin x ) 7 x 6 x x 5 (mod 7) The finite field Euler constant e and the exponential function can be interpreted as follows: (e x ) p [] (e x )p, so that (e x ) p [r ] (e x )p, r e!!! (p )! (mod p) Therefore e i i 4 i 6 i 5i 4 i 5 mod 7 i= 4 5 e i = 4 4 6 6 In order to introduce trigonometric functions, it is possible to consider the complex j since - is a quadratic non-residue for p (mod 4) and pick e j i so that new "e-trigonometric" sin and cos functions are: cos(i)=eexp(ji) 4 i 5i 4 (mod 7), sin(i)=mexp(ji) i 6 i i 5 (mod 7) It is straightforward to verify that cos(i) is even and sin(i) is an odd function The negacyclic Hasse derivative yields: sin [ ] i cos(i) and cos[] i sin(i) We have two kinds of trigonometric functions over a finite field: the k-trigonometric ones and these new e- trigonometric functions Unfortunately, cos(i) and sin(i) do not lie on the unit circle, ie, sin i cos i (mod p) The "number e" is not a "natural" element of the field (in contrast with the real case) However, the k-trigonometric case is defined using an element of the field and presents a lot of properties Nevertheless, derivatives formulae of k- trigonometric functions are not related as usual trigonometric derivatives These results can be generalized so as to define hyperbolic functions over finite fields, p odd, according to the series developments: i GF(p) cosh (i): i! i 4 4! i 6 6! i p (p )! and sinh(i ): i i! i5 5! i7 7! i p (p )! Indeed sinh(i)+cosh(i) exp(i) We recall the convention = ()GF(p) [7] The idea is to consider an extended Galois Field by appending a symbol - as done for the Real number set ([], def9) (x)gf(p): x+(- )- x(- )-and x/(- ) Therefore, other functions such as an hyperbole /x can also be defined i= 4 /i = - 4 Finally, it is interesting to consider "log" functions over a Galois field as the inverse of exponential We have then i= 4 log i = - log () and log () give the expected values MacLaurin series cannot be derived as usual Moreover, the (e x ) function does not have an inverse so (lnx) is not defined, although log x exists, for any primitive element

7 INTEGRAL OVER A FINITE FIELD We evaluate the sum of the n th power of all the p elements of GF(p) It is also assumed that :, since that f()=a (p )? (p )? p p p (p ) p? Lemma (n th power summation) p x n p if n = p - otherwise x proof Letting S n : n n n (p ) n n=,,,,(p-) Trivial case n= apart, n S n S n n S n p n p so that np- implies S S S n (mod p) Definition : The definite integral over GF(p) of a function f from to (p-) is p p I f ( )d : f ( ) Proposition The integral of a function f presenting a series p a(x ) a i x i is given by i p f (x ) mod p p x proof Substituting the series of f in the integral definition and applying Lemma, p p I a i i p p a i x i i i x p a i (p ) i,p i where i,j is the Kronecker delta ( if i=j, otherwise) Corollary If a function f defined GF(p) admits an inverse, then the (p-) th coefficient of its series vanishes The condition a p- mod p is a necessary condition but not a sufficient one to guarantee that f is inversible In fact the integral vanishes since all the images are distinct (the sum of all the elements of the field) Proposition Let f and g be functions defined for every element of GF(p) Denoting by {a i } and {b i } the coefficients of their respective MacLaurin series, then p p p f g ( )d f (k )g (k ) a i b p i p k i proof Substituting the series of f and g in the sum and changing the order of sums, we have p p p f (k )g (k ) a i b j k i j k i,j k The proof follows from Lemma and the fact that the inverse (p-) - p- mod p In order to evaluate the integral over another "interval of GF(p)", we consider: N p N f ()d : f ( ) a i S N (i ), i N where S N (i): x i x A table with the values of S N (i) over GF(5) is showed below i= i= i= i= i=4 S (i) S (i) S (i) 4 S (i) 4 4 S 4 (i) 4 Another interesting result establishes a link between Hasse derivatives and the Finite Field Fourier Transform [] The Galois field function obtained by an image element permutation of f is here referred to as an f- permutation Proposition 4 The coefficients of the McLaurin series expansion of a given signal f are exactly the inverse finite field Fourier transform (FFFT) of an f- permutation Proof Given the -adic development a(x) of a discrete function f, we consider then the finite field Fourier transform pair: a A where the FFFT is from GF(p) to GF(p), and A i =a( i p )= a j ij Since a(x) is used j to interpolate the discrete function f(), for any given i there exist j{-,,,,p-} such that f( j )=a( i )=A i 8 CONCLUSIONS The main point of this paper is to introduce the background of new tools useful for Engineering applications involving finite fields A number of discrete functions is analyzed including a brief look at transcendental functions over Galois Fields Finite field (FF) derivatives are considered, which in turn

leads to Finite Field Taylor Series A brief introduction to FF integration is also presented Clearly many other extensions and applications do exist []- "Transform Techniques for Error-Control Codes", IBM J Res Develop,, n, pp 99-4, May, 979 Acknowledgments- The authors wish to thank Prof VC da Rocha Jr (Universidade Federal de Pernambuco) for introducing us Hasse derivatives References []- W Rudin,"Principles of Mathematical Analysis", McGraw-Hill, 964 []- RJ McEliece, "Finite Fields for Computer Scientists and Engineers", Kluwer Ac Pub, 987 []- RM Campello de Souza, AMP Léo and FMR Alencar, "Galois Transform Encryption", nd Int Symp on Communication Theory and Applications, Ambleside, England, 99 [4]- JL Massey, "The Discrete Fourier Transform in Coding and Cryptography", Info Theory Workshop, ITW, 998, CA: San Diego, Feb 9- [5]- HM de Oliveira, RM Campello de Souza and AN Kauffman, "Efficient multiplex for band-limited channels: Galois-Field Division Multiple Access", WCC'99, Proc of Coding and Cryptography, pp5-4, Paris, France, 999 [6]- RM Campello de Souza, HM de Oliveira, AN Kauffman and AJA Paschoal, "Trigonometry in Finite Fields and a new Hartley Transform", IEEE International Symposium on Information Theory, ISIT, MIT Cambridge,MA, THB4: Finite Fields and Appl, p9,998 [7]- S Lin and DJ Costello Jr, "Error Control Coding: Fundamentals and Applications", NJ: Prentice -Hall, 98 [8]- RW Hamming, "Numerical Methods for Scientists and Engineers", McGraw-Hill, 96 [9]- JL Massey, N von Seemann and PA Schoeller,"Hasse Derivatives and Repeated-root Cyclic Codes", IEEE International Symposium on Information Theory, ISIT, Ann Arbor, USA, 986 []- VC da Rocha, Jr, "Digital Sequences and the Hasse Derivative", Third Int Symp On Communication Theory and Applications, Lake Distric, 95, pp99-6 []- WW Peterson and EJ Weldon Jr, "Error Correcting Codes", Cambridge, Mass: MIT press, 97 []- S Lang, Linear Algebra, Addison-Wesley Pub Co, 966