Internat. J. Math. & Math. Sci. 131 (1987) 131-134 Vol. I0 No. GAPS IN THE SEQUENCE n2 (mod 1) VLADIMIR DROBOT Department of Mathematics Santa Clara University Santa Clara, CA 95053 (Received September 16, 1985) ABSTRACT: Let be an irrational number and let $ For each N let t. [0,I] by I0, I, {k25} the points these intervals can assume. case of the sequence {nS}, KEY WORDS AND PHRASES. I k {t} denote the fractional part of be the intervals resulting from the partition of N 1,2,...,N. Let T(N) be the number of distinct lengths It is shown that T(N) where T(N) <_ 3. This is in contrast to the =o. Uniform distribution mod I. 980 AMS SUBJECT CLASSIFICATION CODE. Primary 1OK31 INTRODUCTION. 1. Let ({t} t fixed N 5 be an irrational number and let {t} denote the fractional part of t (mod I) t It], where [.] is the greatest integer function). For each the points {NS} partition on the interval [0,,] into It is well known that the lengths of these intervals can assume N+I subintervals. only 3 values: {9}, {29}, {3 a, and a+. The values of a and can be actually given explicitly in terms of N and the continued fraction expansion of 9. conjecture and it was first proved by Swierczkowski in exposition of all this, see [2]. This is known as Steinhaus [|]. For an excellent In this note we investigate the analogous problem {n25}. It turns out that in this case the number of different lengths these subintervals can assume, is unbounded. More precisely we have the for the sequence following results. 2. MAIN RESULTS. I be the For each integer N let I0,_I|, N N. k 1,2 N+I subintervals resulting from partition of [0, I] by the points Let T(N) be the number of distinct lengths these subintervals assume. Then for each Theorem Let 5 be an irrational. {k25} > O, T(N) > Nexp {-( +E) in 22 In N In In N for N > N(E) (2 I) V. DROBOT 132 for every 6 > 0 and N > N(6). T(N) >N In particular In what follows $ > 0 is some fixed irrational. We need the following four simple lemmas. LEMMA I. For any integers r,s {r$} {(r+s)$} E where + m (2.2) ;. 0 or We have PROOF. {(r+s)6} (r + s)$ {r$} {r$} {r$} Thus, if 0 < < and if {s$} LEMMA 2. + + [(r+s)$] {uS} {s$} < then (2) holds with E O, + {s} < 2 then (2) holds with E I. {x$} If m If y < x, let x then 0 then e + {k$} {y6} < {x$} {y&} + {k&} {x$} > {y$} {y$}. f {x$} If {(x+y)$} + Then by Lena x + k. contrary to hypothesis, so that E + 0 and (2.3) holds. Again, by Lena m -; so that E {y$} {x$} (2.3) y < x and (2.3) holds again. For any two non-negative integers x,y, 3. PROOF. Then E y + k, k > 0. {x&} E {(x-y)$} {x6} < x < y Suppose x < y so that y {y6} If {x6} + integer (i f } {(y-x)$} PROOF. {s6} + + Suppose x,y are integers, {y$} {r$} + [r] + [s$] then by m m {y$} {x$} # {y$} ee=a 0 or contradicting the fact that $ is irrational. Suppose L 4. {y]$} -{x,$}>0 A If A B then x; II: x x> III: IV: < yl, x < 2 y, y;, y;, < x] x y > x 2, Y2 are non-negative integers and let {y26} -{x25}>0 B x2 Y2 We will use Leman 2 and 3 and consider 4 cases PROOF. I: x;, y, x x x 2 2 2 Y2 > Y2 < Y2 > Y2 In case I we get from Lena 2 A so A {(y] B implies x])$} Y xl B Y2 {(Y2 x2)6} x2 In case II, by Lena 2 we get A {(yl x])$} B {(x 2 Y2)6} GAPS IN THE SEQUENCE so A B cannot hold by Lemma 3. Similarly, A Y2 133 x2 B cannot hold in case III, and A x; in case IV. We are now ready to prove the Theorem I. partition of [0,1] by the points {029} with a collection A(N) of N intervals. [{x9}, {y9}] 2 by Lemma 4. 2 2 xl {N29}. some x), [{x2|9}, If two of these intervals If we we are left {y219 }] and equal length then are of Y; Let N be fixed and consider the {129}, {229}, {329} interval [{x29},I] for 0, exclude the right most interval (i.e. the assume. y; B implies 2 (2.4) x2 Y2 Let T(N) be the number of distinct lengths these intervals from A(N) can The collection A(N) is then divided into T(N) subsets, any two intervals from one subset are of equal length. Thus, by (2 4) k One of these subsets must contain N/T(N) intervals. < k < N 2 such that the equation there exists an integer k, y 2 x 2 (y-x)(y+x) (2.5) has N/T(N) solutions in integers x,y, distinct divisors of k. (xl,y I) (x2,Y2), 2 If then 2 2 < x < y < N 2 Yl xl Y2 x2’ Yl xl Y2 x2 and y| 2 Each such solution produces 2 <- xi + < x < N Yi Y2 + 2 for ,2 and Thus x2" N/T(N) < d(k) (2.6) where d(z) is the number of divisors of d. d(z) < exp{(l+) in 2 In z in in %0( z) z This was first proved by Wigert in [3] It is well known that for each e > 0 see also for z > z(e) [4], Satz 5.2 Since k < N 2 we get from (2.6) 2N/T(N) < %0(e,k) < 0(e,N 2) exp {(I+g) in 2 < exp {(+6) in 2 2 InN (2 7) in 2 + in in N 2 in N in in N for N > N(6) Solving this inequality for T(N) gives (2.1). The argument carries over almost without any change to the sequence integer p > I. THEOREM 2. I0, I], points I {k29}, {nPg} for any The corresponding estimate is then as follows. N k Let 9 be an irrational and p > be the N+| T (N) > N p For each integer N let [0,] by the Let T (N) be the number of distinct lengths these P subintervals resulting from partition of 1,2,...,N. intervals can assume. an integer. Then for each exp{-(l+e) In 2 p > 0 in N In In N for N > N() V. DROBOT 13 REFERENCES SWIERCZKOWSKI, S. On successive settings of an arc on the circumference of a circle, Fund. Math. 46 (1958), 187-189. 2. SLATER, N.B. Gaps and steps for the sequence n$ mod I, Proc. Cambridge Philos. Soc. 63 (1967), Ill5-1123. 3. WIGERT, S. Sur l’ordre de grandeur du numbre des diviseurs d’un entier, Arkiv for Math., Astr., och Fhsik 3, No. 18 (1907), I-9. 4. PRACHAR, K. Primzahlverteilung, Springer Verlag, Gottingen, 1957.