• Aucun résultat trouvé

On linear recurrences with constant coefficients By TRYGVE NAGELL 1.--An arithmetical function A (n)= An of n may be defined by

N/A
N/A
Protected

Academic year: 2022

Partager "On linear recurrences with constant coefficients By TRYGVE NAGELL 1.--An arithmetical function A (n)= An of n may be defined by"

Copied!
7
0
0

Texte intégral

(1)

A R K I V F O R M A T E M A T I K B a n d 3 n r 35

R e a d 13 F e b r u a r y 1957

O n l i n e a r r e c u r r e n c e s w i t h c o n s t a n t coefficients

B y TRYGVE NAGELL

1 . - - A n arithmetical function A ( n ) = An of n m a y be defined b y recursion in the following w a y : The value of An is defined for n = O , 1, 2, .... m - 1, a n d there is given a rule indicating h o w the value of Am+n m a y be determined when the values of A , are known for # = n , n + l , n + 2 . . . n + m - 2 , n + m - 1 , n being a n integer ~ 0.

T h e infinite sequence

A0, A1, A~, A3 ... An ....

thus defined is said to be a recurrent sequence. We denote it b y {A~}. The rule of recursion has often the shape of a recursive /ormula.

F o r instance, t h e function An = n ! satisfies the reeursive formula

A~+I = (n + 1)As (1)

a n d the initial condition A0 = 1.

T h e general solution of t h e reeursive formula

A , + I = 2 A~ (2)

is obviously A~ = 2 ~ A0.

T h e arithmetical function An satisfying the reeursive formula

A~÷~ = V-X~+~ A~

(3)

is a function of n, A0 a n d A1.

Another example is the function An defined b y the recursive formula

A~+2 = An+l + A~ (4)

a n d the initial conditions A 0 = A 1 = 1. I n this case we get the following series:

1, 1, 2, 3, 5, 8, 13, 21 . . . the so-called Fibonacci numbers.

2 . - - I n Algebra and in N u m b e r T h e o r y we often h a v e to do with linear recur- renees, t h a t is to say reeursive formulae of the t y p e

A,~+~ = al Am+n-1 + a2Am+~_e + "" + am A ~ + b, (5) 395

(2)

T. :N'AGELL, On linear recurrences with constant coefficients

where t h e coefficients al, as ... am a n d b are functions of n. I n t h e sequel we shall o n l y consider t h e case in which t h e coefficients are constants. W e shall de- velop a n e l e m e n t a r y t h e o r y of this c a t e g o r y of linear recurrences.

W h e n am # 0 , t h e recurrence (5) is said to be of t h e ruth order. W h e n am =

= a m - l = . . . = a v + l = 0 a n d a ~ # 0 , t h e order of t h e recurrence is/z. I t suffices t o consider t h e case with am # 0 .

T h e r e c u r r e n t sequence

A0, A1, As, As . . . . , An . . . . (6) is said t o be of t h e ruth order if t h e n u m b e r s A , satisfy a linear r e c u r r e n c e of order m b u t no recurrence of a lower order. A r e c u r r e n t sequence of t h e m t h order satisfies e x a c t l y one recurrence of t h e t y p e (5). I n fact, if it satisfied an- o t h e r r e c u r r e n c e

Am+, = cl Am+,-~ + c~Am+n-z + "'" + cm A n + d, we s h o u l d h a v e b y elimination of Am+,

( a l - cl) Am+,-1 + ( a 2 - c2) Am+n-~ + " " + (am - Cm) An + ( b - d) = O.

B u t this recurrence is a t m o s t of order m - 1 .

T h e f o r m u l a (2) is of t h e first order. T h e f o r m u l a (4) is of t h e second order.

W h e n b = 0, t h e recurrence (5) is homogeneous. Tl~e h o m o g e n e o u s recurrence Xm+n = al Xm+n-1. + a2 Xm+n-2 + "'" + am X n (7) is said t o h a v e t h e seals [al, az . . . a m ] .

As a direct consequence of t h e a b o v e definition we h a v e

T h e o r e m t . I ] { A , } a n d {Bn} are two recurrent sequences satislying the recur- rence (7), then { A , + B n } is also a recurrent sequence satisfying (7).

3 . - - W e shall p r o v e

T h e o r e m 2. Suppose that the numbers al, az, as . . . am are given, a m # O.

I f { A n } is a recurrent sequence such that the numbers An satisfy the honmgeneous

r e c u r r e n c e

Am+n = al A m + . - 1 + a2 Am+n-~ + "'" + am An (8) o/ order m, we have the relation

~ A n bo + bl z + b2z 2 + ... + bm_lz m-1

z "

n - o 1 - - a l z - - a 2 z 2 . . . a m Z m ' (9)

where bo, bl, b2 . . . bm-1 are constants which are uniquely determined by A0, A1 . . . Am-l, a l , a ~ , . . . , a m .

Conversely, w h e n bo, bl . . . bin-1 are arbitrarily given constants, the coeHicient8 A , i n (9) satisfy the recurrence (8).

Proof. Given t h e r e c u r r e n t sequence { A , } satisfying (8) it is easy t o see t h a t t h e infinite series

¢o

~ A n z n

(10)

n = 0

(3)

ARKIV FOR MATEMATIK. Bd 3 nr 35 has a certain circle of convergence. I n fact, we will s h o w b y i n d u c t i o n t h a t , for all N_-_O,

I AN I < QN Q~, (11)

where Q = l + l a l [ + l a s l + ... + ] a m ] a n d Qx = m a x ( ] A o [ , I A I ] . . . ]Am_l ]).

T h e relation (11) is clearly t r u e for N = 0 , 1, 2 . . . . , m - 1. I t follows f r o m (8) t h a t (11) is t r u e for N = m . I f we suppose t h a t (11) is t r u e for N = m , m + 1, m + 2 . . . r e + v , we g e t f r o m (8)

[Am+,+x[~_ Q . m a x ( [ A m ÷ , [ . . . [A,+~[) a n d , since (11) is t r u e for all N < = m + v ,

] Am+~+x ] g Q QI" m a x (Qm+,, Qm+,-1 . . . Q~+I) =< Qm+~ Q1.

This proves t h a t (11) is t r u e for all N=>0. H e n c e t h e circle of convergence of t h e series (10) has a radius which is

= l i m s u p ~ 1 _ _ ~ l i m ~ 1 _ 1

Thus, m u l t i p l y i n g t h e series (10) b y t h e p o l y n o m i a l

1 - al z - as z s . . . am z m, (12)

we get, since t h e c o n v e r g e n c e is absolute in t h e inner of t h e circle, t h e following p r o d u c t

m-1

÷n~o a A ~z n+m (13)

~ o b h Z h (A~+n - a l A m + , ~ - l - a 2 A , ~ + m - 2 . . . m ,~, ,

where t h e coefficients ba are u n i q u e l y d e t e r m i n e d b y t h e relations

b o = Ao, ]

L

b 1 = A 1 - a 1 A0,

b s = A s - a 1 A1 - a 2 A 0,

(14)

... j

b i n - 1 = A m - 1 - al A m - s . . . a m - 1 A o . I n v i r t u e of (8) t h e p r o d u c t (13) is equal to t h e p o l y n o m i a l

bo + bl z + bs z ~ + "" + bin-1 Z m -1.

This proves t h e first p a r t of T h e o r e m 2.

Suppose n e x t t h a t t h e n u m b e r s bo, bl . . . bin-1 are a r b i t r a r i l y given a n d e x p a n d t h e rational f u n c t i o n

bo + bl z + b2zg + ... + bm_l Z 'n-1 (15)

1 - - a l z - a s z ~ . . . am z m

(4)

T. NAG]ELL, On linear recurrences with constant coefficients

in a p o w e r series. I f this series is given b y (10), a n d if we m u l t i p l y it b y t h e p o l y n o m i a l (12), we find as a b o v e t h a t t h e coefficients A0, A1, As . . . A a - 1 are d e t e r m i n e d b y t h e s y s t e m (14) a n d f u r t h e r t h a t t h e coefficients A~+~, for all n _>- 0, satisfy t h e r e c u r r e n c e (8).

T h e r a t i o n a l f u n c t i o n (15) is t h e generating /unction of t h e r e c u r r e n t sequence {A=}. This f u n c t i o n m a y be w r i t t e n in t h e f o r m

~o+~, ~+ f l ~ + . . . + ~._~ ~-~

1 - ~ z - o c s z s . . . ~,,z ~ "

where t h e n u m e r a t o r a n d t h e denominafx)r h a v e no c o m m o n divisor z - 0 , a n d where ~ 0 . T h e n t h e order of t h e sequence (An} i s - - / ~ . I n fact, suppose t h a t it was of t h e o r d e r )L<p. T h e n , it would satisfy a h o m o g e n e o u s recurrence w i t h t h e scale [el, cz . . . cA] where e~¢ 0.

H e n c e we s h o u l d have, in v i r t u e of T h e o r e m 2,

flo + f l l z + f l 2 z 2 + . . . + f l , , _ l z ~ - i = e o + e l z + ~ zS + . . . + e x _ i z ~ - 1 1 - ~1 z - ~2 z 2 . . . a ~ z " 1 - - e 1 z - - C 2 Z 2 . . . C A Z ~ '

where Co, e~, ..., ez_~ are constants. T h u s (/~0 +/~1 z + --. + / ~ - 1 z "-~ ) (1 - c l z . . . e ~ ~)

: ( e O " k e l z q - " " q - ex-1 z;t-1) (1 - 0c1z . . . 0~ z~).

H e n c e 1 - cl z . . . ca z a would be divisible b y 1 - ~1 z . . . ~ z ~. B u t this is im- possible since ~ < ~u.

There is of course a n infinity of r e c u r r e n t sequences of a given order m a n d satisfying t h e h o m o g e n e o u s recurrence with t h e given scale [al, a2 . . . am], a~ ¢ 0.

4 . - - W e a d d t h e foUowing result:

T h e o r e m 3. Denote by 01, 02, 0 a, etc. the distinct roots o/ the algebraic equation

Z m - - a l z m - 1 . . . a m - 1 z - am = 0, (16)

where am ~ 0 . T h e n we obtain all the recurrent sequences {An} satis/ying the recur- rence with the scale [al, as . . . am] by the /oUowing formula

o~ t~ v i - 1 + d , ~ - l . ~ \ v ~ _ 2 2 ) + ' " + d 2 . ~ 1

where the s u m is extended over all the distinct roots Ot and where v, is the mul$i- plicity o/ Oz. T h e eoeHicients dl.,, dz., . . . d~, ~ are arbitrary constants.

F o r t h e p r o o f it suffices t o observe t h a t t h e f u n c t i o n (15) m a y be w r i t t e n

[ dvt, i dvt-l.t dl.i ],

~ [ ( 1 - 0~ z) ~ + (1 - 0t z) ~i-1 + " " + 1 -- 0~ zJ a n d t h a t we h a v e t h e expansion

(5)

ARKIV FOR MATEMATIK. Bd 3 Jar 35

( 1 - 0 ~ z ) q . = o O~z ~.

T h e n u m b e r of coefficients dk.~ in f o r m u l a (17) is equal t o m. I f t h e initial values A0, AI . . . Am-1 are given, a n d if the roots of e q u a t i o n (16) are k n o w n , we o b t a i n f r o m (17) a set of m linear equations for t h e d e t e r m i n a t i o n of t h e coefficients dk. ~.

A corollary of T h e o r e m 3 is t h e following p r o p o s i t i o n :

L e t v 1 d e n o t e t h e m n l t i p h c i t y of t h e r o o t 01 of (16), a n d let A= be given b y (17). T h e n t h e difference

B n = A n - d " l \ V1--1 satisfies t h e recurrence

B m + ~ - i = bl Bm+~-2 + b2 Bm+n-3 + "'" + bm-x B . , where t h e coefficients bl, b2 . . . bm_l are d e t e r m i n e d b y t h e i d e n t i t y

Z m - - a i z m - 1 . . . a m : Zm_ 1 _ 51 z m - 2 . . . b i n - 1 . z - 01

W e n o w t u r n t o t h e i n h o m o g e n e o u s recurrences. Consider t h e recurrence o f order m

Am+, = al Am+,-1 + a 2 A , n + , - s + " " + am A n + b, (19) where am a n d b are c 0 . Suppose first t h a t

h = 1 - a l - a 2 . . . a m ¢ O ,

a n d p u t c = b / h a n d Bn = A . - c.

T h e n it is easily seen t h a t B~ satisfies t h e h o m o g e n e o u s recurrence

B m + n = a l B i n + n - 1 + a2 B i n + n - 2 + "'" + a m B n . (20) Suppose n e x t t h a t h = 0. T h e n t h e e q u a t i o n (16) has t h e r o o t z = 1. D e n o t e b y

# t h e multiplicity of this root. W e m a y eliminate b b e t w e e n (19) a n d t h e f o r m u l a

A m + n + l = a l A m + n + a 2 A m + n - 1 + "'" + a m A n + t + b.

T h e n

A m + n + l = (al + 1) Am+. + (as - al) A m + n - 1 + (as - as) A m + n - 2 + " " + + (am -- a m - i ) A n + I - am A n . This recurrence is h o m o g e n e o u s a n d its scale is

[ a l + 1 , a s - - a l , a s - - a2 . . . a m - - a m - l , - - a m ] .

P l a i n l y

Z m + l - - (a i + 1) z m - - ( a s - - al) z m-1 ..... (am -- a m - i ) z + a m

= (z-- 1) (z m -- a l z m-1 -- a ~ z m-2 . . . am).

(6)

T. NAGELL, On linear recurrences w i t h constant coefficients

Hence, in virtue of the corollary of Theorem 3, we h a v e (for all n >_-0)

A , = B , + ( n ~ / z ) d ,

(21)

where B . satifies (20), a n d where d is a certain constant (i. e. independent of n).

T o determine d we h a v e the relations

Am+, = Bm+, + (m + n + /zl d

a n d

ai Am+,~-t = ~ a~ Bm+,,-i + t as d.

t f f i l i f f i l

Since Am a n d B . satisfy (19) a n d (20) respectively, we get b y subtraction

b= m~-n+/z --_as #

Since b a n d d are constants, we m a y take n = 0 . Hence

d=

b

# / i-1 #

where /z denotes the multiplicity of the root z = 1 in equation (16).

I n this w a y the inhomogeneous case has been reduced to t h e homogeneous case.

5 . - - A general t h e o r y of recurrences is developed in N. E. NSrlund,

Vorl~ungen giber Dif/erenzrechnung,

Berlin 1924 (Verl. Springer). Linear recurrences are t r e a t e d in K a p i t e l 14, § 2. B u t the m e t h o d employed is quite different f r o m the simple one a d o p t e d in this note.

6 . - - W e finish with a few examples:

(1) The sequence of t h e

Fibonar~i numbers F,~

1, 1, 2, 3, 5, 8, 13, 21 . . . . satisfy the recurrence of the second order

F . + 2 =

F.+I

+

F~

a n d the initial conditions F 1 = F 2 = 1. One finds easily

(2) The general solution of the recurrence of the first order A , + l = a A n + b

400

(7)

ARKIV FOR MATEMATIK. B d 3 n r 3 5

n " a n - 1 .

is easily found to be A . = a A0 + ~ 0, valid also for a = 1.

(3) The coefficients A . in the expansion

1 ~ A

1 - 7 z ~ - 6 z a .=o satisfy the recurrence of the third order

A,~+s= 7 A,~+x + 6 An,

a n d the initiM conditions A 0 = l , A I = 0 , A ~ = 7 . Further we find A . = - 1 ( - 1)n + ~ ( - 2)" +~o3".

I n fact the equation z 3 - 7 z - 6 = 0 has the roots z = - 1 , - 2 , and 3.

(4) I f we p u t in the reeursive formula (3), for all n, B . = log A . ,

we get the homogeneous linear recurrence B,~+2=I B.+I + ~ B . .

Hence B . = (x + fl ( _ ~)1 n,

where a a n d /3 arc determined b y the relations

B0= ~+/3, B1=~--~/3.

Bn=a gBo + gBl + ~ ( B o - B1) ( - ~ ) n 2 Thus

a n d finally

A 3 _ A 2 ( - ½ ) n + l / ] 2 - 2 ( - ½)n

T r y c k t d e n 15 m a r s 1957

U p p s a l a 1957. Almqvist & Wiksslls B o k t r y c k e r i A B

Références

Documents relatifs

Parashar and Choudhary [27] have intro- duced and examined some properties of four sequence spaces defined by us- ing an Orlicz function M, which generalized the well-known

Given a relation which is first order definable in Presburger arith- metic of Z (resp. N) it is recursively decidable whether or not it is first order definable in the structure

In the continuous settings, we restrict ourselves to the treatment of (real func- tions over) Euclidean spaces R k with their natural topology. From the viewpoint of

determining a certain configuration S in Euclidean n-dimen- sional space; in what cases and by what methods can one pass from the equation to the topological

In the sequel, we explain that even though γ is conjecturally not a value at an algebraic point of these arithmetic special functions, certain representations hidden in (1)(15)

The definition of the spectrum given by Berkovich is the following: Let (k, |.|) be complete field with respect to an ultrametric absolute value and let A 1,an k be the Berkovich

Furthermore, if a definition of negation in terms of incompatibility helps itself to the notions of truth and falsity, one might as well define negation right away through

This paper provides an effective method to create an abstract simplicial complex homotopy equivalent to a given set S described by non-linear inequalities (polynomial or not)1. To